Building a Prediction Model for Radiographically Confirmed Pneumonia in Peruvian Children: From Symptoms to Imaging

A Deep Learning Model for the Diagnosis and Discrimination of Gram-Positive and Gram-Negative Bacterial Pneumonia for Children Using Chest Radiography Images and Clinical Information

Purpose

This study aimed to develop a deep learning model based on chest radiography (CXR) images and clinical data to accurately classify gram-positive and gram-negative bacterial pneumonia in children to guide the use of antibiotics.

Methods

We retrospectively collected CXR images along with clinical information for gram-positive (n=447) and gram-negative (n=395) bacterial pneumonia in children from January 1, 2016, to June 30, 2021. Four types of machine learning models based on clinical data and six types of deep learning algorithm models based on image data were constructed, and multi-modal decision fusion was performed.

Results

In the machine learning models, CatBoost, which only used clinical data, had the best performance; its area under the receiver operating characteristic curve (AUC) was significantly higher than that of the other models (P<0.05). The incorporation of clinical information improved the performance of deep learning models that relied solely on image-based classification. Consequently, AUC and F1 increased by 5.6% and 10.2% on average, respectively. The best quality was achieved with ResNet101 (model accuracy: 0.75, recall rate: 0.84, AUC: 0.803, F1: 0.782).

Conclusion

Our study established a pediatric bacterial pneumonia model that utilizes CXR and clinical data to accurately classify cases of gram-negative and gram-positive bacterial pneumonia. The results confirmed that the addition of image data to the convolutional neural network model significantly improved its performance. While the CatBoost-based classifier had greater advantages owing to a smaller dataset, the quality of the Resnet101 model trained using multi-modal data was comparable to that of the CatBoost model, even with a limited number of samples.

The utility of chest x-ray and lung ultrasound in the management of infants and children presenting with severe pneumonia in low-and middle-income countries: A pragmatic scoping review

Background

Chest x-ray (CXR) is commonly used (when available) to support clinical management decisions for child pneumonia and provide a reference standard for diagnosis in research studies. However, its diagnostic and technical limitations for both purposes are well recognised. Recent evidence suggests that lung ultrasound (LUS) may have diagnostic utility in pneumonia. This systematic scoping review of research on the utility of CXR and LUS in the management of severe childhood pneumonia aims to inform pragmatic guidelines for low- and middle-income countries (LMICs) and identify gaps in knowledge.

Methods

We included peer-reviewed studies published between 2000 and 2020 in infants and children aged from one month to nine years, presenting with severe pneumonia. CXR studies were limited to those from LMICs, while LUS studies included any geographic region. LUS and CXR articles were mapped into the following themes: indications, role in diagnosis, role in management, impact on outcomes, and practical considerations for LMIC settings.

Results

85 articles met all eligibility criteria, including 27 CXR studies and 58 LUS studies. CXR studies were primarily observational and examined associations between radiographic abnormalities and pneumonia aetiology or outcomes. The most consistent finding was an association between CXR consolidation and risk of mortality. Difficulty obtaining quality CXR images and inter-reader variability in interpretation were commonly reported challenges. Research evaluating indications for CXR, role in management, and impact on patient outcomes was very limited. LUS studies primarily focused on diagnostic accuracy. LUS had higher sensitivity for identification of consolidation than CXR. There are gaps in knowledge regarding diagnostic criteria, as well as the practical utility of LUS in the diagnosis and management of pneumonia. Most LUS studies were conducted in HIC settings with experienced operators; however, small feasibility studies indicate that good inter-operator reliability may be achieved by training of novice clinicians in LMIC settings.

Conclusions

The available evidence does not support the routine use of CXR or LUS as essential tools in the diagnosis and initial management of severe pneumonia. Further evaluation is required to determine the clinical utility and feasibility of both imaging modalities in low-resource settings.

Radiology for Thoracic Conditions in Low- and Middle-Income Countries

With a disproportionately high burden of global morbidity and mortality caused by chronic respiratory diseases (CRDs) in low and middle-income countries (LMICs), access to radiological services is of critical importance for screening, diagnosis, and treatment guidance.

Effectiveness of the 10-valent pneumococcal conjugate vaccine on pediatric pneumonia confirmed by ultrasound: a matched case-control study

Background

Bangladesh introduced the 10-valent pneumococcal conjugate vaccine (PCV10) for children aged < 1 year in March 2015. Previous vaccine effectiveness (VE) studies for pneumonia have used invasive pneumococcal disease or chest X-rays. None have used ultrasound. We sought to determine the VE of PCV10 against sonographically-confirmed pneumonia in three subdistrict health complexes in Bangladesh.

Methods

We conducted a matched case–control study between July 2015 and September 2017 in three subdistricts of Sylhet, Bangladesh. Cases were vaccine-eligible children aged 3–35 months with sonographically-confirmed pneumonia, who were matched with two types of controls by age, sex, week of diagnosis, subdistrict health complex (clinic controls) or distance from subdistrict health complex (community controls) and had an illness unlikely due to Streptococcus pneumoniae (clinic controls) or were healthy (community controls). VE was measured using multivariable conditional logistic regression.

Results

We evaluated 8926 children (average age 13.3 months, 58% boys) with clinical pneumonia by ultrasound; 2470 had pneumonia with consolidations ≥ 1 cm; 1893 pneumonia cases were matched with 4238 clinic controls; and 1832 were matched with 3636 community controls. VE increased with the threshold used for consolidation size on ultrasound: the adjusted VE of ≥ 2 doses vs. non-recipients of PCV10 against pneumonia increased from 15.8% (95% CI 1.6–28.0%) for consolidations ≥ 1 cm to 29.6% (12.8–43.2%) for consolidations ≥ 1.5 cm using clinic controls and from 2.7% (− 14.2–17.2%) to 23.5% (4.4–38.8%) using community controls, respectively.

Conclusions

PCV10 was effective at reducing sonographically-confirmed pneumonia in children aged 3–35 months of age when compared to unvaccinated children. VE increased with the threshold used for consolidation size on ultrasound in clinic and community controls alike. This study provides evidence that lung ultrasound is a useful alternative to chest X-ray for case–control studies evaluating the effectiveness of vaccines against pneumonia.

Chest radiography in children aged 2-59 months enrolled in the Innovative Treatments in Pneumonia (ITIP) project in Lilongwe Malawi: a secondary analysis

BACKGROUND

Pneumonia is the leading infectious cause of death in children aged under 5 years in low- and middle-income countries (LMICs). World Health Organization (WHO) pneumonia diagnosis guidelines rely on non-specific clinical features. We explore chest radiography (CXR) findings among select children in the Innovative Treatments in Pneumonia (ITIP) project in Malawi in relation to clinical outcomes.

METHODS

When clinically indicated, CXRs were obtained from ITIP-enrolled children aged 2 to 59 months with community-acquired pneumonia hospitalized with treatment failure or relapse. ITIP1 (fast-breathing pneumonia) and ITIP2 (chest-indrawing pneumonia) trials enrolled children with non-severe pneumonia while ITIP3 enrolled children excluded from ITIP1 and ITIP2 with severe pneumonia and/or selected comorbidities. A panel of trained pediatricians classified the CXRs using the standardized WHO CXR research methodology. We analyzed the relationship between CXR classifications, enrollee characteristics, and outcomes.

RESULTS

Between March 2016 and June 2018, of 114 CXRs obtained, 83 met analysis criteria with 62.7% (52/83) classified as having significant pathology per WHO standardized interpretation. ITIP3 (92.3%; 12/13) children had a higher proportion of CXRs with significant pathology compared to ITIP1 (57.1%, 12/21) and ITIP2 (57.1%, 28/49) (p-value = 0.008). The predominant pathological CXR reading was "other infiltrates only" in ITIP1 (83.3%, 10/12) and ITIP2 (71.4%, 20/28), while in ITIP3 it was "primary endpoint pneumonia"(66.7%, 8/12,; p-value = 0.008). The percent of CXRs with significant pathology among children clinically cured (60.6%, 40/66) vs those not clinically cured (70.6%, 12/17) at Day 14 was not significantly different (p-value = 0.58).

CONCLUSIONS

In this secondary analysis we observed that ITIP3 children with severe pneumonia and/or selected comorbidities had a higher frequency of CXRs with significant pathology, although these radiographic findings had limited relationship to Day 14 outcomes. The proportion of CXRs with "primary endpoint pneumonia" was low. These findings add to existing data that additional diagnostics and prognostics are important for improving the care of children with pneumonia in LMICs.

TRIAL REGISTRATION

ITIP1, ITIP2, and ITIP3 were registered with ClinicalTrials.gov ( NCT02760420 , NCT02678195 , and NCT02960919 , respectively).

Ten Years of Pediatric Lung Ultrasound: A Narrative Review

Lung diseases are the most common conditions in newborns, infants, and children and are also the primary cause of death in children younger than 5 years old. Traditionally, the lung was not thought to be a target for an ultrasound due to its inability to penetrate the gas-filled anatomical structures. With the deepening of knowledge on ultrasound in recent years, it is now known that the affected lung produces ultrasound artifacts resulting from the abnormal tissue/gas/tissue interface when ultrasound sound waves penetrate lung tissue. Over the years, the application of lung ultrasound (LUS) has changed and its main indications in the pediatric population have expanded. This review analyzed the studies on lung ultrasound in pediatrics, published from 2010 to 2020, with the aim of highlighting the usefulness of LUS in pediatrics. It also described the normal and abnormal appearances of the pediatric lung on ultrasound as well as the benefits, limitations, and possible future challenges of this modality.

Design and Comparative Performance of a Robust Lung Auscultation System for Noisy Clinical Settings

Chest auscultation is a widely used clinical tool for respiratory disease detection. The stethoscope has undergone a number of transformative enhancements since its invention, including the introduction of electronic systems in the last two decades. Nevertheless, stethoscopes remain riddled with a number of issues that limit their signal quality and diagnostic capability, rendering both traditional and electronic stethoscopes unusable in noisy or non-traditional environments (e.g., emergency rooms, rural clinics, ambulatory vehicles). This work outlines the design and validation of an advanced electronic stethoscope that dramatically reduces external noise contamination through hardware redesign and real-time, dynamic signal processing. The proposed system takes advantage of an acoustic sensor array, an external facing microphone, and on-board processing to perform adaptive noise suppression. The proposed system is objectively compared to six commercially-available acoustic and electronic devices in varying levels of simulated noisy clinical settings and quantified using two metrics that reflect perceptual audibility and statistical similarity, normalized covariance measure (NCM) and magnitude squared coherence (MSC). The analyses highlight the major limitations of current stethoscopes and the significant improvements the proposed system makes in challenging settings by minimizing both distortion of lung sounds and contamination by ambient noise.

Real-World Verification of Artificial Intelligence Algorithm-Assisted Auscultation of Breath Sounds in Children

Objective: Lung auscultation plays an important role in the diagnosis of pulmonary diseases in children. The objective of this study was to evaluate the use of an artificial intelligence (AI) algorithm for the detection of breath sounds in a real clinical environment among children with pulmonary diseases. Method: The auscultations of breath sounds were collected in the respiratory department of Shanghai Children's Medical Center (SCMC) by using an electronic stethoscope. The discrimination results for all chest locations with respect to a gold standard (GS) established by 2 experienced pediatric pulmonologists from SCMC and 6 general pediatricians were recorded. The accuracy, sensitivity, specificity, precision, and F1-score of the AI algorithm and general pediatricians with respect to the GS were evaluated. Meanwhile, the performance of the AI algorithm for different patient ages and recording locations was evaluated. Result: A total of 112 hospitalized children with pulmonary diseases were recruited for the study from May to December 2019. A total of 672 breath sounds were collected, and 627 (93.3%) breath sounds, including 159 crackles (23.1%), 264 wheeze (38.4%), and 264 normal breath sounds (38.4%), were fully analyzed by the AI algorithm. The accuracy of the detection of adventitious breath sounds by the AI algorithm and general pediatricians with respect to the GS were 77.7% and 59.9% (p < 0.001), respectively. The sensitivity, specificity, and F1-score in the detection of crackles and wheeze from the AI algorithm were higher than those from the general pediatricians (crackles 81.1 vs. 47.8%, 94.1 vs. 77.1%, and 80.9 vs. 42.74%, respectively; wheeze 86.4 vs. 82.2%, 83.0 vs. 72.1%, and 80.9 vs. 72.5%, respectively; p < 0.001). Performance varied according to the age of the patient, with patients younger than 12 months yielding the highest accuracy (81.3%, p < 0.001) among the age groups. Conclusion: In a real clinical environment, children's breath sounds were collected and transmitted remotely by an electronic stethoscope; these breath sounds could be recognized by both pediatricians and an AI algorithm. The ability of the AI algorithm to analyze adventitious breath sounds was better than that of the general pediatricians.

Challenges in the diagnosis of paediatric pneumonia in intervention field trials: recommendations from a pneumonia field trial working group

Pneumonia is a leading killer of children younger than 5 years despite high vaccination coverage, improved nutrition, and widespread implementation of the Integrated Management of Childhood Illnesses algorithm. Assessing the effect of interventions on childhood pneumonia is challenging because the choice of case definition and surveillance approach can affect the identification of pneumonia substantially. In anticipation of an intervention trial aimed to reduce childhood pneumonia by lowering household air pollution, we created a working group to provide recommendations regarding study design and implementation. We suggest to, first, select a standard case definition that combines acute (≤14 days) respiratory symptoms and signs and general danger signs with ancillary tests (such as chest imaging and pulse oximetry) to improve pneumonia identification; second, to prioritise active hospital-based pneumonia surveillance over passive case finding or home-based surveillance to reduce the risk of non-differential misclassification of pneumonia and, as a result, a reduced effect size in a randomised trial; and, lastly, to consider longitudinal follow-up of children younger than 1 year, as this age group has the highest incidence of severe pneumonia.

Training and standardization of general practitioners in the use of lung ultrasound for the diagnosis of pediatric pneumonia

BACKGROUND

Pneumonia is a leading cause of death in children of low-resource settings. Barriers to care include an early and accurate diagnosis. Lung ultrasound is a novel tool for the identification of pediatric pneumonia; however, there is currently no standardized approach to train in image acquisition and interpretation of findings in epidemiological studies. We developed a training program for physicians with limited ultrasound experience on how to use ultrasound for the diagnosis of pediatric pneumonia and how to standardize image interpretation using a panel of readers.

METHODS

Twenty-five physicians participating in the training program conducted lung ultrasounds in all children with suspected pneumonia, aged 3 to 35 months, presenting to three subdistrict hospitals in Sylhet, Bangladesh, between June 2015 and September 2017.

RESULTS

A total of 9051 pediatric lung ultrasound assessments were conducted through 27 months of data collection. Study physicians underwent training and all were successfully standardized, achieving 91% agreement and maintained a sensitivity and specificity of 88% and 92%, respectively, when their diagnosis was compared with experts. Overall kappa between two readers was high (0.86, 95% confidence interval [CI], 0.84-0.87), and remained high when a third expert reader was included (0.80, 95% CI, 0.79-0.81). Agreement and kappa statistics were similarly high when stratified by age, sex, presence of danger signs, or hypoxemia.

CONCLUSIONS

Lung ultrasound is a novel tool for the diagnosis of pediatric pneumonia with evidence supporting its validity and feasibility of implementation. Here we introduced a training program that resulted in a high level of inter-sonographer agreement.

Radiologic Diagnosis and Hospitalization among Children with Severe Community Acquired Pneumonia: A Prospective Cohort Study

Objectives

This study was designed to assess the role of chest radiography for the diagnosis of pneumonia and assess the association of clinical characteristics with radiologic findings and predictors of hospitalization among children with severe community acquired pneumonia.

Methods

A prospective study was conducted on 122 children between ages of 3 month and 14 years admitted to pediatric emergency unit with diagnosis of severe pneumonia from September 1^st to November 30^th, 2017. Eligible children were subjected to chest radiography which was read by two senior radiologists independently (R1 and R2). Disagreements between R1 and R2 were resolved by a third senior radiologist (R3). Level of agreement between radiologists was assessed using Cohen's kappa coefficient. Clinical and laboratory parameters which could explain the variability in the duration of hospital stay were assessed using a linear regression mode. Independent predictors were assessed using multiple linear regression.

Results

The median age of the cohort was 10.0 months (interquartile range (IQR): 6.75-24.0); 76 (62.3%) were male. Nearly half, 63 (51.6%) did not have radiologic evidence of pneumonia. There was low level of agreement between R1 and R2 in reporting consolidation (kappa=0.435, p-value≤0.001), haziness (kappa=0.375, p-value≤0.001), and infiltration (kappa=0.267, p-value=0.008). Children with higher recorded temperature were more likely to have radiologic abnormalities suggesting pneumonia (p-value=0.033). The median duration of hospitalization was 3 days (IQR: 1-4 days); 118 (96.7%) were discharged with improvement. Height-for-age z-score (Coef.=0.203, R²=0.041, p-value=0.027); and hemoglobin level (Coef.=-0.249, R²=0.062, p-value=0.006) explained 4.1% and 6.2% of the variability in the duration of hospital stay, respectively.

Conclusion

Radiologic evidence of pneumonia was absent in half of the children with severe pneumonia. There was low agreement between senior radiologists in reporting chest radiographic findings, potentially necessitating harmonization activities to uniformly implement the WHO guidelines in reading chest radiographs.